Contributions to Colon Segmentation Without Previous Preparation in Computer Tomography Images

Abstract

We propose a method to segment the lumen of the colon from computed tomography (CT) images. To do so, we use first and second order statistical moments. These moments provide us with a set of descriptors to characterize the homogeneity of regions inside the colon. The algorithm presented in this paper makes use of these values in a prediction-correction exploration process of the colonic region. We show some segmentation results on real patient images that present high nonhomogeneous colonic regions.